A Former Market Maker’s Perception of PFOF

Payment for order flow, market segmentation, and returns to scale

It feels like payment for order flow controversies flare up every few years. When I see some of the takes I know how marine biologists felt after Jaws hit the cinemas in 1975.

Except they didn’t have Twitter to scream into.

I’m going to assume you already know what payment for order flow is.

If you need the basics, A16’s Alex Rampell and Scott Kupor have you covered. (Link)
If you want the GOAT of high finance’s version, here is the Matt Levine post I shared last week. (Link)

Now if you stop at Levine’s post I’d forgive you. There’s really no following that guy. But now that I’ve said that, you own the downside of reading further and if I say anything useful here I’m in-the-money.

I think my experience qualifies me to hopefully add some perspective to the discussion. I have been trading options for 21 years with the first half of those years on the floor. Even though I’ve been trading prop for the past decade I’m a dyed-in-the-wool market-maker. You can take the dog off the floor, but you can’t take the floor out of the dog. (Full disclosure: I used to work for SIG who was an early payer for order flow, but I had no insight into that side of their business).

An Image Problem

Payment for order flow sounds terrible. It sounds like payola. Greasing the radio DJ to get your record played on-air. That’s a bribe to the regional gatekeeper. There’s widespread misconception that when Citadel pays for flow it’s attempting to use the info to front-run the order. This is a dizzying misconception.

No trader thinks front-running random retail flow makes any sense.

Write that on a chalkboard 50x please.

The Nature of Adverse Selection

Drive it home: no trader thinks front-running random retail flow makes any sense.

In fact, the opposite is true. The entire basis of trading against retail flow is that it is a random mix of buys and sells and not autocorrelated. You want to trade against your drunk uncle Sal who has a good feeling about the Jets this Sunday. We call this “dumb” flow. Sorry, but that’s what it’s called.

On the other hand, we refer to institutional flow as “smart flow”. Not because it knows which direction the stock is going to go, although this can be the case as anyone who has been contra to SAC flow back in the day can attest. The reason we don’t want to trade against the flow is that it’s autocorrelated. 1,000 shares is the tip of an iceberg. Nobody eats just one chip just as nobody buys just 1,000 shares.

The options equivalent is putting someone up on a trade, only to have them reload 5 minutes later. This past fall, Softbank string-raised tech calls every day for a couple weeks. Masa-son is not smart paper, but he has a big stack. Truthfully, the threshold to be an undesirable counterparty is surprisingly low. I remember hearing a SIG trader at a conference a few years after I left. He mentioned that their studies had shown that the adverse selection of an options trade went up dramatically once it was greater than 16 lots.

Let’s understand this. Consider a pro-rata exchange where your limit bid is on equal standing with other limit bids but your fill is proportional to your size. So let’s say you are bidding $1.25 for 100 option contracts and the total bid quantity is 1000. If a retail sized order sells the bid for 10 contracts, you get filled on 1 because your size was 10% of the total displayed size. The pro-rata system (vs maker-taker which is a queue based on speed) incentivizes traders to show far more liquidity than they really want to. They don’t want to get their whole bid hit but they need to show size to be entitled to any reasonable percentage of the incoming orders. When an order sweeps the book, banging out the displayed size on the bid, the market makers are instantly sad. They know they are on the wrong side of a “smart” order.

The possibility that the flow you trade against is adverse, smart, institutional – whatever you want to call it – has a deep implication. You make a wider market than you would have if you could just tell the difference between the adverse flow and the random retail flow.

THIS IS THE EUREKA MOMENT WHEN INSIGHT RUSHES IN…

The brokers have realized they can segment the market between orders that can be facilitated on tighter spreads and those that require wider quotes. Liquidity has a price. Without PFOF, spreads need to be cushioned by the probability that an order is institutional. Instead, PFOF creates a tiered market where the cost of liquidity is proportionally aligned with the risk on a per trade basis. Retail traders get better fills. There’s less deadweight loss.

Institutional traders might complain, but its an illusion that they should have gotten the price that a retail trader should get. The risk business is not the widget business. You don’t get volume discounts.


Analogies 

“The opportunity to trade against random flow” as a source of revenue is a bit abstract. You are already familiar with price discrimination in other domains.

  • Casino’s attracting whales
    Casinos don’t like card counters, they want customers that have positive LTV in the long run. They like whales and the type of people who buy books titled “The Fool-Proof System To Beating Roulette”. Casinos are paying for order flow when they offer complimentary suites and blacked out SUVs to and from McCarran.
  • Ad tech
    What is the internet but reams of data on customers being sold to the highest bidder so platforms (the brokers in our analogy) and in turn vendors (the Citadels) more can more efficiently convert sales (trades)?
  • Financial products
    Good driver discounts on auto policies. Life insurance physicals. Credit checks for loans. Price discrimination based on risk is the norm not the exception.
  • Retail
    As a broke 20 year old I used to frequently buy and return products at GNC. Yes, you can return a half-used tub of creatine. GNC started keeping tabs as a policy. I get it. The Ponderosa wishes it could turn away Joey Chestnut.

Competition

The discourse around PFOF has an air of monopoly sentiment. Maybe not in the Standard Oil sense of the world. There’s more firms than Citadel. You have Virtu, G1 (SIG), Two Sigma, Wolverine. It looks more like OPEC.

But there’s a big difference. These are not natural monopolies or crony handouts. Contrast the dynamic with payola. Payola was a scam that worked because the value of the bribe to the briber (the record label) was very low compared to the payoff of getting radio exposure. Meanwhile the value of the bribe was substantial to the receiving DJ who was paid a conventional salary despite being the caretaker of a government monopoly — airwaves.

I don’t think it’s surprising that high fixed cost industries settle into oligopoly-type hierarchies. The competitive forces are so strong that they double as high barriers to entry. The HFT-firms here are not defending natural monopolies. They are the survivors of the trading game who invested heavily in technology early. @hidenotslide explains in his recent post about another storied traded firm, DRW:

This brings me to my first point – firms who embraced HFT early in its evolution are today’s kings. Of the 10-20 firms that make up the bulk of high frequency trading profits, a large majority were launched before the 2008 financial crisis and many even prior to 2000. Because superior technology leads to direct competitive advantages in HFT, barriers to entry have become insurmountable over the last decade as companies have invested in ever faster exchange connections & market data feeds. A 2017 paper from researchers at Cornell & Penn argues this exact point – newer, smaller entrants that engage in HFT can survive, but they don’t get anywhere near the share of profits that larger, more established firms enjoy.

What’s absent from the narrative is how tall the pile of bodies these firms stand atop. I should know. I used to be able to work five hours a day (NYMEX alum holla) and make a lawyer’s wage. And in some years, a law partner’s carry too. Well, if you were smart you saved your money and realized it wasn’t going to last. The days of “locals” (ie wildcat market-makers) is long gone.

Many of the small firms, who saw the writing on wall and had an appetite for the long game, plowed money back into massive technology capex. Most of them just earned the right to say they lost to the best. In some cases they found small, profitable niches where they play the role of suckerfish. Respect to them, even this was not easy.

How about the remaining firms? The private giants the media likes to call “shadowy”. They were the ones who were most adept at assembling teams of software and hardware engineers working with game-theory geniuses to devise algos in a cat-and-mouse battle with competitors. The ones who stayed step-for-step with the exchanges who themselves were experimenting with matching engine rules, data, product listings and connectivity in their own battles for market share.

The truth is progress is cutthroat.

I remember the days before decimalization where you could make $5 wide verticals 3/8 wide. Today that same vertical is a choice market and the market maker gets paid the equivalent of an inter-dealer broker commission or about 25 cents. On a 3/8 wide market the market maker used to earn nearly $18.75 (or 50% of 3/8)! My business partner and I always marvel at the innovation and how little vig a trader is willing to accept to flip million dollar coins. It’s such a flex for capitalism. So much so that how good these firms are is chalked up to monopoly and not that fact that they are the survivors of the capitalism’s most brutal tournament.

How Survivorship Bias Makes Firms Look Like Monopolies

Perhaps I should not be surprised at the monopoly sentiment. Some of you will nod. “How can they make money every day?” First, I’m not sure they do, but even if they did that’s hardly a red flag. Casinos might make money every day so long as they can open. They’re not monopolies. Worrying that financial firms make money everyday is conflating market makers with investment managers because they traffic in the same products. But one of them is a customer and the other is a supermarket. With tiny supermarket margins per trade. And high fixed costs. If volumes dried up, the losses would show up even if the margins stayed flat.

A stronger, but still naïve argument, that they were monopolies would come from noticing that these shops came of age at the same time as the giant tech firms. This is a hint of how much they have in common. The difference is the size of the relative opportunities, but the tactics are similar.

It started with skill and luck. The early big bets on talent and technology meant they were bringing guns to a knife fight. SIG wasn’t know as the “evil empire” on the Amex just because of the black jackets we wore. They understood the risk-reward was completely outsized to what it should be 25 years ago. They were amongst the first to tighten markets to steal market share. They accepted slightly worse risk-reward per trade but for way more absolute dollars. They then used the cash to scale more broadly. This allowed them to “get a look on everything”. Which means you can price and hedge even tighter. Which means you can re-invest at a yet faster rate. Now you are blowing away less coordinated competitors who were quite content to earn their hundreds of percent a year and retire early once the markets got too tight for them to compete.

SIG was playing the long game. The parallels to big tech write themselves. A few firms who bet big on the right markets start printing cash. This kicks off the flywheel:

Provide better product –> increase market share –> harvest proprietary data. Circle back to start.

The lead over your competitors compounds. Competitors die off. They call you a monopoly.

Equilibriums

Thus far I’ve only pushed back against the idea that PFOF is somehow nefarious. It is a form of price discrimination. The price discrimination is economically sensible when we price liquidity. There is a cost to having someone trade with you at the exact moment you want to trade. If you are a retail trader, that cost is tiny and we can thank technology and the competitive drive of very smart people to undercut one another so they can be the best bid for your business.

If you are an institutional trader that cost is higher. And it should be. Your cost to trade should be compared to your historical cost to trade. Not against what a retail trader’s costs are. I’d be shocked if an apples-to-apples TCA showed that this cost has increased over time. My null is the cost to trade for everyone has collapsed but probably more for retail.

I don’t have any strong opinions as to whether PFOF is the best equilibrium. One could argue we should have a single central order book, but then the exchange would have a monopoly. Plus it’s not obvious to me that the centralization of liquidity serves the heterogenous interests of all economic stakeholders across countries, regulatory regimes, strategies, time zones, and instruments.

We could entertain more incremental tweaks to the current architecture. For example an auction every minute or shorter trading hours to centralize liquidity in time but not venue. There’s probably some efficient frontier of tradeoffs. Nothing about PFOF looks villainous from my understanding of markets so if it lies along that frontier I would not be surprised.

And perhaps now you won’t be either.